Semiparametric Estimation of Partially Varying-Coefficient Dynamic Panel Data Models

Abstract

This paper studies a new class of semiparametric dynamic panel data models, in which some of coefficients are allowed to depend on other informative variables and some of the regressors can be endogenous. To estimate both parametric and nonparametric coefficients, a three-stage estimation method is proposed. A nonparametric generalized method of moments (GMM) is adopted to estimate all coefficients firstly and an average method is used to obtain the root-N consistent estimator of parametric coefficients. At the last stage, the estimator of varying coefficients is obtained by the partial residuals. The consistency and asymptotic normality of both estimators are derived. Monte Carlo simulations are conducted to verify the theoretical results and to demonstrate that the proposed estimators perform well in a finite sample.